Quantum-assisted NMR Inference on Noisy Rydberg Atom Devices

Inferring molecular parameters from NMR spectra, an important task in the biomedical sciences, is a computationally challenging task. Hybrid quantum-classical approaches on near-term devices could make NMR inference tractable. A recent proposal couples quantum computation of the NMR spectra of hypothetical molecules with machine learning to optimize molecular parameters of small molecules [1]. With an eye towards implementing this hybrid approach, we study Hamiltonian simulation on NISQ Rydberg atom devices. Specifically, we construct an experimentally realistic error model for Rydberg atoms and use it to optimize protocols which balance hardware error with Trotterization error to achieve high-fidelity Hamiltonian simulation for spectral computation. On the learning side, we study the proliferation of spurious local minima in the NMR inference landscape of an exactly solvable model. Our approach provides an analytical lens into the barren plateaus problem for NMR inference.

[1] Sels, D., Dashti, H., Mora, S. et al. Quantum approximate Bayesian computation for NMR model inference. Nat Mach Intell 2, 396–402 (2020)